You detect an isolated 52 Hz axial room mode with a measured T60 of 0.82 s. On REW's Pole-Zero plot, how does a properly configured Modal filter neutralize this resonance compared to an arbitrary narrow peaking boost/cut?
Lesson 05 • Filter Synthesis, Modal Damping & Complex-Plane Verification
Lesson 5 — EQ, Modal Analysis and Target Matching
Move beyond blind auto-flattening. Use REW's EQ window as a time-frequency filter synthesis environment: couple room-curve and house-curve target architecture with modal T60 estimation, strict boost/Q constraints, headroom accounting, equaliser-accurate biquad modeling, and pole-zero verification.
01 • Engineering Rationale
What this is for
In room calibration, loudspeakers and listening environments form a mixed-phase system: direct-radiation driver transfer functions and isolated low-frequency room modes behave closely to minimum-phase systems, whereas boundary cancellations, comb filtering, and early specular interference introduce non-minimum-phase zeroes. A casual EQ pass that simply forces a steady-state SPL trace onto a flat horizontal line degrades transient accuracy, injects high-Q ringing, destroys digital headroom, and fails catastrophically the moment the listener moves 15 cm off-axis.
REW's EQ window solves this by unifying six coupled engineering domains inside a single workspace:
- Hardware-accurate filter modeling: The Equaliser panel locks parameter ranges, step resolution, Q/bandwidth definitions, filter topologies, and sample rate to your actual DSP target (miniDSP, CamillaDSP, rePhase, StormAudio, Hypex, Dutch & Dutch 8c, Generic/Extended, etc.) so predicted traces match hardware biquads rather than idealized math.
- Acoustically realistic target construction: Separate natural loudspeaker/subwoofer roll-off and bass-management crossover slopes (Target Settings) from subjective tonal balance (Room Curve LF rise / HF fall or external House Curve files), ensuring auto-EQ never fights driver physics.
- Time-domain modal damping: Modal Analysis identifies individual room resonances and estimates their decay time (T60), enabling Modal filters whose Q is specifically calculated to reduce ringing to a target decay time—not merely shave steady-state amplitude.
- Constraint-governed optimisation: Match Range, Individual/Overall Max Boost, Boost Q limits, flatness targets, and the narrow-filter frequency boundary (< 200 Hz) keep automated matching strictly inside minimum-phase, spatially stable regions.
- Full complex-domain prediction: Because REW convolves the measured impulse/transfer function with the candidate filter chain, Predicted Waterfall, Predicted Impulse, Predicted Phase, Predicted Group Delay, and the Pole-Zero plot expose non-minimum-phase artifacts before a single coefficient is flashed to hardware.
- Headroom governance: The EQ Filters panel continuously computes cumulative filter gain and required headroom so positive filter summation never drives downstream fixed-point DSP, DACs, or amplifiers into clipping.
EQ is the last corrective tool in the signal chain. Complete physical placement, acoustic treatment, subwoofer/main time-and-polarity alignment (Alignment Tool), and variable smoothing first. Never use parametric filters to compensate for a crossover cancellation notch, a gross channel-level offset, or a measurement captured without a reliable IR window.
02 • Deterministic REW Workflow
Steps
Execute this sequence in order. Reversing steps—such as matching before selecting the equaliser model or before establishing target level—invalidates Q constraints, filter quantization, and headroom projections.
-
Prepare the Measurement Trace & Smoothing
Select the validated single-position or spatially averaged measurement in the main window. Apply Variable smoothing (Graph → Variable smoothing or Cmd+8 on macOS): this preserves unsmoothed 1/48-octave resolution below 100 Hz for accurate modal identification while progressively smoothing to 1/3 octave above 1 kHz so the optimiser does not chase high-frequency comb filtering. Confirm the measurement sample rate matches your target DSP clock.
-
Open the EQ Window & Lock the Equaliser Model
Click EQ in the main toolbar. Open the Equaliser panel on the right and select the exact Manufacturer and Model (e.g., miniDSP 2x4 HD, CamillaDSP, rePhase, Hypex, StormAudio, Dutch & Dutch 8c, or Generic / Configurable PEQ). Verify the sample rate shown in the Equaliser panel title—biquad coefficients and high-frequency frequency warping depend directly on this rate.
-
Configure Target Architecture (Speaker Physics First)
Open Target Settings and choose the Speaker Type matching the acoustic DUT:
- Full Range: No imposed LF/HF crossover target beyond the room/house curve.
- Bass Limited: Set LF Cutoff and LF Slope (e.g., 12 dB/oct sealed or 24 dB/oct vented) to match the speaker's natural acoustic roll-off so REW does not boost below f3.
- Subwoofer: Configure both LF Cutoff/Slope (driver/box subsonic limit) and Bass Management low-pass crossover slope/cutoff (e.g., 24 dB/oct Linkwitz-Riley at 80 Hz).
- Speaker Driver: Configure dedicated high-pass/low-pass acoustic crossover target shapes when designing active multi-way crossovers.
-
Layer the Room Curve (LF Rise / HF Fall) or Load a House Curve
Inside Target Settings, define the desired in-room tonal tilt using either:
- Built-in Room Curve: Enable Add room curve and set LF Rise Start (e.g., 100–200 Hz), LF Rise End (e.g., 20–30 Hz), and LF Rise Slope (dB/octave), plus HF Fall Start (e.g., 1 kHz) and HF Fall Slope (typically 0.5–1.0 dB/octave depending on room directivity and listening distance).
- External House Curve File: Load a plain-text file (via Preferences → House Curve or Target Settings) containing whitespace-separated frequency offset_dB pairs. REW linearly interpolates in log-frequency space between points and sums the offset onto the speaker target.
-
Set Target Level Deliberately (Protect Headroom)
Click Calculate target level from response to establish a baseline, then inspect the target trace against the measured curve across your match band. Adjust Target Level manually downward by 1–3 dB if necessary so the target sits at or below the average passband SPL—prioritizing cuts on peaks rather than forcing broadband boosts. Never use EQ filters to correct a master calibration offset between channels; fix trim gain in the processor first.
-
Run Modal Analysis & Quantify Low-Frequency Resonances
Open the Modal Analysis panel. Enter room length, width, and height to overlay theoretical axial/tangential/oblique mode lines. Configure Search Range (typically 20–250 Hz), Analysis Length, and Noise Threshold, then click Find Resonances. Inspect the Resonances panel listing each detected mode's exact Frequency, SPL, and estimated T60. Cross-reference these peaks against the Excess Group Delay and Waterfall graphs to confirm minimum-phase modal behavior.
-
Define Match Range, Boost Limits, Flatness & Narrow-Filter Guardrails
Open Filter Tasks and lock down the optimiser constraints before running automatic matching:
- Match Range: Set to the modal region (e.g.,
20–200 Hzor up to Schroeder frequency fs) for room-mode correction, or a wider band only when applying gentle, low-Q tonal shaping based on a spatially averaged response. - Individual Max Boost & Overall Max Boost: Set to
0 dBfor strict cut-only modal EQ, or at most1–3 dBwith tight Q caps if filling a verified minimum-phase broad depression. - Flatness Target: Typically
1–2 dB; overly tight tolerances (e.g., 0.5 dB) exhaust filter banks chasing irrelevant ripples. - Allow narrow filters below 200 Hz: Enable for modal peaks; leave high-Q filtering disabled above the transition region.
- Allow shelves / Vary max Q above 200 Hz: Enable shelves when matching broad tonal tilt; keep boost Q limits conservative so the optimiser never synthesizes high-Q artificial resonances.
- Match Range: Set to the modal region (e.g.,
-
Synthesize Filters, Assign Modal Filters & Fine-Tune in EQ Filters
Click Match Response to Target. Open the EQ Filters panel to audit every generated biquad (Type, Control state, Frequency, Gain, Q/Bandwidth). For prominent isolated low-frequency modes identified in Step 6, manually change the filter type from
PKto Modal and specify the measured resonance T60 and desired target T60 (or fine-tune interactively on the Filter Adjust graph). Use Optimise gains, Optimise gains and Qs, or Optimise gains, Qs and Fcs after locking manual modal filters (uncheck their auto-match control checkbox so optimisation respects them). -
Verify Time-Domain, Phase, Pole-Zero & Headroom Predictions
Switch across the EQ window graph tabs—Waterfall (toggle live update if needed), Impulse, Phase, Group Delay, and Pole-Zero—comparing Measured vs Predicted traces. Check the Headroom readout in the EQ Filters panel. Confirm that predicted modal ridges collapse cleanly, excess group delay does not spike, no zeroes/poles create instability, and required headroom is accommodated by input attenuation in your DSP.
-
Deploy or Audition via Hardware Export / Measure with These Filters
In Filter Tasks, either:
- Measure with these filters: REW applies the modelled filter chain directly to the outgoing sweep signal, letting you acoustically verify linearity, distortion, and room interaction before programming external hardware.
- Save filter coefficients to file: Export sample-rate-specific biquad coefficients (miniDSP, SigmaStudio, etc.).
- Export filter settings as text: Generate structured parameter tables for CamillaDSP, rePhase, Equalizer APO, or system documentation.
- Send filter settings to equaliser: Transfer over network (e.g., Dutch & Dutch 8c) or supported comms interface—then always capture a post-EQ verification sweep.
On macOS, keyboard modifiers map to Command (Ctrl) and Option (Alt). Direct hardware transfers over RS232 serial (such as TAG McLaren AV32R DP / AV192R) are Windows 32-bit only. MIDI port selection is disabled by default on macOS because Java MIDI access can crash the JRE on some Macs; if communicating with legacy MIDI hardware (BFD Pro DSP1124P / FBQ2496), enable MIDI cautiously in Preferences → Comms, or prefer file-based biquad/text export and network transfer where supported.
03 • Diagnostic Interpretation
Reading the output
Every graph and panel inside the EQ window answers a distinct physical question. Reading only the steady-state Predicted magnitude trace hides phase distortion, inadequate damping, and non-minimum-phase violations.
Filter Adjust View — Target, Constraints & Headroom
Filter AdjustFilter Adjust anatomy: With Individual/Overall Max Boost = 0 dB and Match Range bounded to 20–200 Hz, REW targets the minimum-phase 49 Hz modal peak and 135 Hz hump while refusing to waste power or inject ringing into the 86 Hz boundary cancellation notch or alter above-Schroeder directivity.
Pole-Zero Plot — Modal Resonance Cancellation
Pole-ZeroPole-Zero interpretation: Angular position encodes frequency; radial distance r to the unit circle encodes damping (T60). A properly tuned Modal filter cancels a near-unit-circle resonance pole with a coincident zero and substitutes a deeper pole with shorter decay.
Panel-by-Panel & Graph-by-Graph Reference
| EQ Window Panel / Graph | What It Displays | Professional Diagnostic Check |
|---|---|---|
| Filter Adjust | Measured SPL, Predicted SPL, Target, Filters response, and Filters+Target traces with direct graphical drag-editing of filter Fc, gain, and Q. | Verify Predicted converges on Target across Match Range without deep secondary lobes. Confirm Filters trace returns cleanly to 0 dB outside the corrected band unless a deliberate shelf is active. |
| EQ Filters | Active filter bank: Enable, Control (Auto/Manual), Type (PK, Modal, LS/HS, LP/HP, notch, etc.), Frequency, Gain, Q or Bandwidth, plus cumulative Headroom indicator. |
Check for redundant overlapping filters (two adjacent PK filters fighting each other), unintended positive-gain filters consuming headroom, and hardware-quantised parameter snaps. |
| Modal Analysis & Resonances | Room dimension mode overlay, search range, analysis length, noise threshold, and tabular list of detected resonances (Frequency, SPL, T60). | Distinguish isolated high-T60 axial modes (prime candidates for Modal filters) from overlapping cluster peaks or low-SNR noise artifacts below the threshold. |
| Waterfall (Predicted) | Measured vs Predicted 3D spectral decay (supports optional live updates during filter tweaking). | Confirm that cutting a minimum-phase modal peak actually shortens its time-domain ridge. If the steady-state peak drops but a late ridge persists unchanged, the resonance is poorly coupled or non-minimum-phase at that mic position. |
| Impulse / Phase / Group Delay | Measured vs Predicted impulse response, phase trace, and group delay. | Because REW's predictions retain non-minimum-phase components, watch for group-delay spikes introduced by excessive boost Q or high-order crossovers. Properly damped minimum-phase modal peaks will visibly flatten predicted group delay. |
| Pole-Zero | Complex-plane positions of resonance poles (×) and filter poles/zeroes (× / ○) relative to the unit circle. | Check pole proximity to the unit circle: a high-Q boost filter places its own pole close to |z| = 1, synthesizing artificial ringing. A modal cut pulls net pole radius inward, verifying time-domain stabilization. |
| Target Settings | Speaker Type, LF cutoff/slope, Bass Management crossover, Room Curve (LF rise / HF fall), House Curve file, and Target Level. | Confirm LF cutoff matches natural driver roll-off so REW does not attempt to force sub-cutoff excursion; confirm Target Level sits below passband mean if pursuing zero-boost calibration. |
| Filter Tasks | Match Range, Individual/Overall Max Boost, Flatness Target, shelf & <200 Hz narrow-filter toggles, Match/Optimise actions, and Export/Send/Measure triggers. | Audit constraints before every auto-match run. Re-running Optimise gains and Qs after locking manual Modal filters often trims 2–3 redundant biquads while improving overall RMS error. |
Good vs. Pathological EQ Signatures
- Clean modal damping signature: A 45–60 Hz peak drops 8–12 dB in Predicted SPL, its Waterfall ridge retreats by 150–300 ms, Predicted Group Delay flattens across the mode, and the Pole-Zero plot shows the modal filter zero overlaid on the resonance pole with the replacement pole deeper inside the unit circle.
- Headroom-neutral target tracking: Target Level is anchored slightly below passband average; filters are exclusively cuts (or ≤ +2 dB low-Q shelving); EQ Filters headroom remains 0 dB loss (or explicitly offset by matched input attenuation in DSP).
- High-Q boost ringing artifact: Attempting to fill a 10 dB SBIR or room-mode cancellation notch with a +9 dB, Q=12 PK filter creates a filter pole perilously close to the unit circle—Predicted Waterfall shows a new long-decaying ridge at the notch frequency, Headroom plunges by 9+ dB, and off-axis listeners suffer a severe narrowband boom.
- Sub-cutoff excursion trap: Setting Target Type to Full Range on a ported monitor with a 48 Hz roll-off (or allowing boosts below natural roll-off) drives massive low-frequency boost into the unloaded port region, risking driver over-excursion and gross harmonic distortion.
- Above-Schroeder narrowband chasing: Allowing high-Q filters above 200–300 Hz on an unsmoothed single-mic measurement corrects localized comb filtering valid only inside a few cubic centimeters, introducing audible tonal coloration everywhere else in the listening zone.
04 • Advanced Engineering Theory
Deeper Detail • DSP Mathematics, Equaliser Physics & Automation Pole-Zero Modal Math, Filter Q Conventions, Headroom Budgeting & API Workflows
1. Pole-Zero Geometry & Why Modal Filters Beat Generic PK Cuts
An isolated room mode behaves locally like a second-order resonator whose transfer function contains a complex conjugate pole pair at radius rm and angle ω0 = 2πf0/fs. Because an exponentially decaying sinusoid decays as rmn per sample, the time required to fall by 60 dB (10−3 in amplitude) maps directly to the pole radius:
r_m = exp( -ln(1000) / (T60 * f_s) ) ≈ exp( -6.9078 / (T60 * f_s) )
Q_m ≈ (pi * f_0 * T60) / 6.9078 ≈ 0.455 * f_0 * T60
For example, a 50 Hz mode ringing with T60 = 0.85 s exhibits an intrinsic modal quality factor Qm ≈ 19.3. A standard peaking filter (PK) chosen solely to flatten steady-state magnitude can easily have the wrong zero radius: if its Q is too wide, it over-attenuates adjacent frequencies; if its Q is too narrow, it leaves skirts of the modal pole uncanceled.
REW's Modal filter inverts the problem directly in pole-zero space: given the resonance frequency f0, measured T60meas, and desired T60target (or gain), REW places the biquad's numerator zeroes directly on top of the room's poles (rz = rm) to cancel the long-ringing denominator, and positions the biquad's denominator poles radially inward at rp corresponding to T60target. Because pole-zero cancellation requires exact spatial and frequency alignment, Modal filters are most effective on well-separated, minimum-phase axial modes below ~150 Hz and must be validated across nearby listening positions.
2. Non-Minimum-Phase Zeroes & Why Predicted Phase Tells the Truth
When a direct wave sums with a delayed, attenuated boundary reflection of opposite phase, the resulting notch corresponds to a transfer-function zero located outside the unit circle (|z| > 1) in the z-plane. Inverting that notch requires placing an IIR filter pole at the same exterior coordinate—which is mathematically unstable (impulse response grows exponentially to infinity). Any causal, stable minimum-phase EQ boost applied at that frequency can only boost the total energy (direct + reflected) at that exact coordinate while injecting massive excess group delay and ringing.
Crucially, REW does not assume the measured room is minimum-phase when computing EQ predictions. It multiplies the complex measured transfer function (including all excess phase) by the complex filter response. Consequently, Predicted Phase, Predicted Group Delay, Predicted Impulse, and Predicted Waterfall remain physically faithful: minimum-phase cuts cancel both magnitude and excess ringing, whereas non-minimum-phase distortions remain stubbornly visible in the predictions.
3. Equaliser Models, Q vs. Bandwidth Conventions & Biquad Quantisation
"Q = 4.0 at -6 dB" does not produce the identical transfer function across every DSP platform. Manufacturers define filter bandwidth differently:
- RBJ / Constant-Q vs. Symmetric / Proportional-Q: Some equalisers maintain constant -3 dB (or half-gain) bandwidth regardless of boost/cut depth, while proportional-Q designs narrow bandwidth as gain increases.
- Q vs. Octave Bandwidth (BW): Legacy hardware units (e.g., BFD DSP1124P, FBQ2496) express width in fractions of an octave (1/60 oct steps) rather than Q, with coarse gain/frequency quantisation grids.
- Sample-Rate Bilinear Warping: Biquad coefficients (b0, b1, b2, a1, a2) exported via Save filter coefficients to file (miniDSP, SigmaStudio) are strictly locked to the sample rate displayed in the Equaliser panel title. Loading 48 kHz biquads into a 96 kHz DSP shifts every centre frequency up by one full octave!
- Model-Specific Constraints: Always select the exact hardware profile (e.g., miniDSP 2x4 HD at 96 kHz, CamillaDSP, rePhase, StormAudio, Hypex, Dutch & Dutch 8c) before matching. Note hardware quirks: BFD DSP1124P requires Store Enable to persist transferred presets, and FBQ2496 MIDI transfers may require increasing inter-filter delay if packets drop.
4. Headroom Budgeting, Filter Summation & Gain Structure
The Headroom figure in the EQ Filters panel reports the peak positive gain across frequency of the combined filter bank. However, professional DSP gain staging must account for two additional realities beyond steady-state magnitude headroom:
- Phase-Induced Crest Factor Growth: Even an all-cut (0 dB peak magnitude) IIR filter chain alters phase relationships among harmonics of a broadband transient (or full-scale square-like program material), which can increase time-domain peak levels by 1–3 dB. Always verify peak levels in Predicted Impulse and leave at least 3 dB digital attenuation upstream of EQ in fixed-point or near-0 dBFS pipelines.
- House Curve Headroom Cost: A +6 dB LF rise at 20 Hz relative to 1 kHz consumes 6 dB of headroom if implemented as positive boost. Instead, shift the Target Level downward (or apply -6 dB global preamp attenuation before the EQ block) so the peak of the house curve sits at 0 dB filter gain and the midrange/HF are gently attenuated—preserving digital ceiling and amplifier linearity.
5. Multi-Position Spatial Averaging vs. Single-Point EQ
Because a transfer function is strictly valid only for one source and one microphone position, generating high-Q EQ above the modal band from a single sweep optimizes a microscopic sweet spot. For multi-seat or studio console calibrations, capture 5–9 positions across the listening envelope, inspect spatial stability in All SPL, compute an RMS average (incoherent magnitude average, avoiding phase-cancellation artifacts of unaligned vector averages), and run target matching on the averaged curve—while keeping individual-position modal verification sweeps handy to confirm no seat develops a resonant peak.
6. Programmatic EQ Inspection & Automation via REW API (macOS)
For repeatable multi-channel or production-line workflows, REW exposes local REST endpoints on localhost:4735 (including /eq for defaults, equalisers, target settings, house curves, and match settings, plus /measurements for per-measurement EQ and filters). Launch REW on macOS with the API enabled:
open -a REW.app --args -api
curl -s http://127.0.0.1:4735/eq/defaults | python3 -m json.tool
GET methods inspect EQ configurations and measurement filter banks immediately; PUT/POST automated control and sweep execution require the REW Pro upgrade. The API binds strictly to 127.0.0.1 for security.
05 • Applied Calibration Case Study
Worked example: Dual-Goal Modal Damping & Custom House Curve Matching
Calibrate a subwoofer + bass-limited studio monitor system (already time- and polarity-aligned at an 80 Hz LR24 crossover) in a 4.8 × 3.6 × 2.7 m control room on macOS at 48 kHz. Specifically: (1) isolate and damp a severe axial room mode near 48.6 Hz using a T60-targeted Modal filter, (2) apply a subtle +4.5 dB LF rise / -0.6 dB/oct HF fall house curve, (3) maintain zero boost (0.0 dB headroom loss) inside the 20–200 Hz modal match band, and (4) verify both frequency and time-domain performance before exporting CamillaDSP / miniDSP filter parameters.
| Parameter | Configured Value | Engineering Justification |
|---|---|---|
| Sample Rate & DSP Target | 48 kHz • Generic / CamillaDSP (or miniDSP 2x4 HD at 96 kHz if applicable) |
Matches interface clock in macOS Audio MIDI Setup; prevents biquad frequency warping. |
| Input Trace Conditioning | Variable smoothing (full-range aligned L+Sub measurement) | Retains 1/48-oct resolution across 20–100 Hz modes while smoothing HF comb ripples. |
| Target Type & LF Limit | Bass Limited • LF Cutoff: 24 Hz, Slope: 24 dB/oct |
Tracks natural vented-sub acoustic roll-off; blocks sub-cutoff boost and over-excursion. |
| House Curve / Room Curve | External house curve: +4.5 dB @ 25 Hz → 0 dB @ 120 Hz → -3.0 dB @ 20 kHz | Delivers warm, non-fatiguing in-room tilt without forcing broadband positive EQ gain. |
| Match Range & Boost Caps | 20–200 Hz • Individual Max Boost: 0 dB • Overall Max Boost: 0 dB |
Confines high-Q correction to minimum-phase modal region; guarantees cut-only headroom safety. |
| Modal Filter Target | 48.6 Hz mode (measured T60 = 0.78 s) → Target T60 = 0.28 s | Cancels axial length/width resonance pole directly instead of blind magnitude flattening. |
Step-by-Step Execution
-
Create and Load the House Curve File
Save the following plain-text file as
studio-house-curve.txtand load it in Preferences → House Curve (or directly inside EQ → Target Settings):studio-house-curve.txt (Frequency Hz Offset dB)20 4.5 25 4.5 60 2.2 120 0.0 1000 0.0 10000 -2.4 20000 -3.0 -
Set Target Architecture & Anchor Target Level
In EQ → Equaliser, select
CamillaDSP(confirming 48 kHz in panel title). In Target Settings, set Speaker Type toBass Limited, LF Cutoff =24 Hz, LF Slope =24 dB/oct. Verify the loaded house curve appears summed into the dashed Target trace. Click Calculate target level from response (yields ~76.2 dB SPL), then nudge Target Level down to 74.5 dB SPL so the target tracks the lower envelope of the modal passband without demanding boost. -
Execute Modal Analysis & Lock the 48.6 Hz Modal Filter
Open Modal Analysis, enter room dimensions (
4.80 m × 3.60 m × 2.70 m), set Search Range20–200 Hz, and click Find Resonances. The Resonances panel reports a dominant mode at 48.6 Hz (+11.2 dB above target, T60 = 0.78 s), matching the room's second-order width / tangential cluster. Open EQ Filters, set Filter 1 to TypeModal, Frequency48.6 Hz, Measured T600.78 s, Target T600.28 s(yields approx. -9.8 dB cut, equivalent Q ≈ 10.6), and set its control checkbox to Manual so subsequent auto-matching preserves it. -
Run Constrained Match & Optimise Remaining Modal Band
Open Filter Tasks. Set Match Range =
20–200 Hz, Individual Max Boost =0 dB, Overall Max Boost =0 dB, Flatness Target =1.5 dB, check Allow narrow filters below 200 Hz, and uncheck Allow shelves (the speaker's broad bass balance already tracks the house curve). Click Match Response to Target, then run Optimise gains and Qs. -
Inspect Predicted Graphs, Pole-Zero & Acoustically Verify
Inspect Filter Adjust, Predicted Waterfall, Predicted Group Delay, and Pole-Zero. Click Measure with these filters to run a live validation sweep through the modelled filter chain without leaving REW, then export via Export filter settings as text (or Save filter coefficients to file for miniDSP).
Expected Output
| Filter # | Control | Type | Freq (Hz) | Gain (dB) | Q / T60 | Acoustic Purpose |
|---|---|---|---|---|---|---|
| 1 | Manual | Modal |
48.6 | -9.8 | T60: 0.78 → 0.28 s | Cancels dominant axial/tangential pole & collapses ringing |
| 2 | Auto | PK |
71.4 | -5.2 | Q = 6.8 | Tames 2nd length axial mode (~71.5 Hz in 4.8 m room) |
| 3 | Auto | PK |
114.2 | -4.1 | Q = 4.9 | Reduces vertical/width upper modal cluster |
| 4 | Auto | PK |
158.0 | -2.6 | Q = 3.2 | Smooths upper modal transition before 200 Hz cutoff |
Interpretation & Verification Checkpoints
- Filter Adjust: Measured +11.2 dB peak at 48.6 Hz is pulled cleanly onto the +2.8 dB house-curve target contour. A localized -7 dB dip at 89 Hz (front-wall SBIR cancellation) remains completely unboosted because Max Boost = 0 dB prevents the optimiser from attacking a non-minimum-phase zero.
- Waterfall & Pole-Zero: The 48.6 Hz decay ridge shortens from ~780 ms to ~310 ms in Predicted Waterfall (and confirms within ±40 ms on the live Measure with these filters verification sweep). On the Pole-Zero plot, Filter 1's zero sits squarely on the 48.6 Hz resonance pole and relocates the damped pole safely inward.
- Headroom: The EQ Filters panel reports 0.0 dB required headroom (pure cut-only filter bank). Applying a conservative -2.0 dB master pre-gain in CamillaDSP easily absorbs any minor phase-rotation crest-factor rise on hot masters.
- Spatial Check: Re-measuring ±30 cm left/right of the mix position confirms the 48.6 Hz and 71.4 Hz cuts improve decay uniformity across the entire console width, because both modes are spatially coherent pressure maxima across that zone.
06 • Knowledge Verification
Check your understanding
Four scenario-based diagnostic questions covering modal pole-zero cancellation, non-minimum-phase limits, target/headroom architecture, and equaliser deployment fidelity.
A subwoofer measurement shows a sharp -11 dB notch at 94 Hz accompanied by a localized spike in Excess Group Delay. If you override REW's constraints to apply a +10 dB, Q = 11 peaking filter at 94 Hz, what will REW's Predicted views and physical room response reveal?
You are matching a bass-limited monitor (natural acoustic f3 = 46 Hz, 24 dB/oct vented roll-off) to a house curve with a +5 dB LF rise. Which Target Settings practice prevents both driver over-excursion and unnecessary digital headroom loss?
You synthesize EQ filters in REW on macOS with the Equaliser model set to Generic (48 kHz), export raw biquad coefficients via Save filter coefficients to file, and import them into a DSP engine running at 96 kHz. What is the direct consequence?